In Focus delivers deeper coverage of the political, cultural, and ideological issues shaping America. Published daily by senior writers and experts, these in-depth pieces go beyond the headlines to give readers the full picture. You can find our full list of In Focus pieces here.
Five years after the pandemic-era census miscounted hundreds of thousands of people across more than a dozen states, Republicans are still grappling with the implications of the errors — and plotting how to ensure the next census doesn’t repeat them.
The Trump administration is eyeing ways to republish census data that could aid in mid-decade redistricting. Republicans in Congress are demanding an investigation into how the U.S. Census Bureau mishandled the 2020 count. And conservatives are pushing to include a citizenship question on the next census that could lead to the exclusion of millions of illegal immigrants from the official population totals used to dole out representatives to states.
The 2020 census overcounted the populations of eight states and undercounted them in six states. All but one of the states that were undercounted were red states, and all but two of the states that were overcounted were blue states.
Critics have also pointed to a new statistical formula the Census Bureau used to analyze the 2020 census data that scrambled the distribution of population within states and their demographic numbers in ways that may have distorted congressional maps.
“All of those – by chance, by design, I can’t tell you what the answer is — they tended in the aggregate to benefit only Democrats when it came to actual apportionment,” Wade Miller, senior adviser at the Center for Renewing America, told the Washington Examiner.
Because census data is constitutionally required to determine the number of U.S. House seats, Electoral College votes, and the amount of federal funding each state receives, the stakes of the decennial survey are enormously high. Republicans in Congress are advancing legislation that would block the Census Bureau from using federal funds to count illegal immigrants for the purposes of apportionment, among other reforms that could significantly alter the 2030 survey.
But Democrats are fiercely opposed to changes that could shift congressional seats out of their reach. Setting aside possible fixes to the census, they are already facing that prospect: Americans have moved out of blue states such as California and New York in droves since the pandemic and settled primarily in red states, all but ensuring that the 2030 apportionment process will add to the ranks of red state delegations until at least 2040.
Pandemic problems
The 2020 census’s errors were far more significant than those that occurred in 2010.
The Post-Enumeration Survey, which the Census Bureau conducts after every census to gauge the accuracy of the results, found that the 2020 census more or less accurately tallied the total population of the United States but significantly miscounted the populations of 14 states — in some cases, by hundreds of thousands of people.
By contrast, after the 2010 census, the Post-Enumeration Survey “did not measure a statistically significant undercount or overcount in the population or housing units for any state.” In total, the PES for the 2010 census found that just 36,000 people were overcounted.
After the 2020 census, the PES found that more than 695,000 people were overcounted in New York alone.
“Those are significant errors, and that’s why folks are still talking about it today,” Hans von Spakovsky, manager at the Heritage Foundation’s Election Law Reform Initiative, told the Washington Examiner. “The states that got overcounted are getting more money than they should, and the states that got undercounted … are getting cheated.”
SHUTDOWN SHOWS REPUBLICANS NEED A HEALTHCARE SOLUTION
When presenting the results of the PES in 2022, Census Bureau officials blamed the pandemic for hindering the count.
Then-Census Bureau Director Robert Santos said the agency faced an “unprecedented set of challenges” conducting the massive survey amid lockdowns and the spread of COVID-19. While he acknowledged that there were “areas of concern” within the data, he argued the numbers were still “fit for use” in the political process.
Erika Becker-Medina, then chief of the Decennial Communications Coordination Office, said the pandemic forced the bureau to pause its in-person field operations in the spring of 2020.
“Under normal census circumstances, we would have completed in-person field operations before hurricane season was in full swing,” Becker-Medina said. “Instead, we hit our peak operations as the nation faced multiple devastating hurricanes.”
Fear of the virus also affected in-person counting operations because, Becker-Medina explained, people were more hesitant than usual to interact with “strangers at their door.”
“We changed our procedures to limit in-person contact with the public,” she said. The Census Bureau also extended its data collection window by two and a half months to try to make up ground lost during stay-at-home orders.
But the pandemic can’t explain all of the problems with the final results. The 2020 census was the first to offer an online response option, and millions of Americans took advantage of it. Sixty-seven percent of households self-reported either online or by mailing in a paper response before the October 2020 deadline. And while Florida was among the states that faced an erroneous population count, most of the states that saw overcounts or undercounts would have been unaffected by the active hurricane season the Atlantic saw that year.
Statistical sleights of hand
Some Republicans have questioned whether a new statistical formula used by the Census Bureau, in theory to protect respondent privacy, could have further distorted the 2020 numbers.
Known as differential privacy, the algorithm was designed to inject what’s known as “noise” into the data to prevent anyone in possession of it from being able to identify individual respondents at the most granular level. The algorithm, used for the first time in the presentation of the 2020 census data, did this by randomly assigning some demographic data from respondents into different geographic categories to conceal precisely who is living where.
Imagine the census collection process as a series of small cups pouring water into bigger cups that then pour water into buckets that, eventually, all get dumped into a giant pool. The smallest cups in the census are known as blocks, and they all must have, on average, the same amount of water in them before they are dumped in the census tract cups, then into county cups, and eventually, into the state buckets. Differential privacy allowed Census Bureau analysts to randomly distribute the water in the smallest cups before pouring them into the next-biggest cups, rather than just directly transferring the water they collected.
This did not affect the total amount of water that ended up in the state bucket, or even in the national pool. But critics say the algorithm resulted in an inaccurate picture of how populations were actually distributed across the state, which legislatures then used to draw their congressional maps.
“As problematic as the 2020 census was for apportionment, it may have been disastrous for redistricting,” Sen. Jim Banks (R-IN) wrote in a letter to Commerce Secretary Howard Lutnick earlier this month. “Differential privacy alters the total population of individual voting districts, meaning that any voting district drawn since the 2020 census may in fact be based on false data and may even be unlawful.”
The full extent of the problem remains unknown, Banks noted, because the raw 2020 census data was never released to the states and remains under lock and key at the Census Bureau.
A 2021 Harvard University analysis, which described the decision to use differential privacy in the census as “controversial,” found that the use of the algorithm likely resulted in census data that was “not of sufficient quality for redistricting purposes.”
By applying a version of the algorithm to 2010 data, Harvard’s test showed the method was “biased against certain areas” and that it “makes it impossible for states to accurately comply with the One Person, One Vote principle.” The idea of “One Person, One Vote” stems from a 1964 Supreme Court decision that said every congressional district must have roughly the same number of people in it.
“This was likely disproportionately impacting rural communities,” said Miller of the Center for Renewing America. “You need to inject more noise into areas with smaller populations to make it harder for the algorithm to be reverse engineered, and you need to do it less in heavily populated areas.”
“Although it would not necessarily impact apportionment, it could impact the way political districts are drawn, because you’re drawing maps based off of fake population data,” he added.
Re-emergence of redistricting
Some Republican states, egged on by President Donald Trump, have begun the process of mid-decade redistricting as questions about the 2020 census draw renewed attention.
But those states can only shuffle around the number of seats given to them by the results of the 2020 census, regardless of any mistakes discovered in the years since it was conducted.
“There’s nothing in federal law that would allow a state to sue the federal government and the Census Bureau, saying, ‘Hey, you admitted you undercounted me,’” said von Spakovsky of the Heritage Foundation. “There’s no legal remedy for it.”
Florida should have gotten two additional congressional seats that it did not receive due to undercounting in the 2020 census, von Spakovsky’s analysis found. Texas would have gotten an extra seat if not for its undercount, and the overcounts in both Rhode Island and Minnesota meant each got to keep a seat it should have lost, he said.
Conservatives say the census errors overall cost the GOP at least six congressional seats Republicans would otherwise have held this decade.
And they hope to secure even more representation by ensuring only U.S. citizens are counted in the numbers that determine apportionment.
Citizenship question questioned
Trump tried and failed during his first term to include a citizenship question on the 2020 census, but the Supreme Court ruling that emerged from the legal battle over the issue did not preclude him from trying again.
Instead, the Supreme Court decided in 2019 that the Trump administration had not provided a coherent reason for including a citizenship question on the census, blocking it on those procedural grounds but leaving the door open for inclusion if they were resolved.
“The Enumeration Clause permits Congress, and by extension the Secretary [of Commerce], to inquire about citizenship on the census questionnaire,” the Supreme Court said, citing the “long history of the citizenship question on the census.”
The citizenship question was included in all but one census between 1820 and 2000 in at least some part of the survey — either the one posed to all households or a longer form that only some households receive. But in 2010, during the Obama administration, the Census Bureau removed the question altogether and instead included it in what’s known as the American Community Survey, a data collection effort the Census Bureau conducts each year that does not affect apportionment.
Republicans in Congress have introduced several proposals to address the issue of illegal immigration in the 2030 census, including one that would subtract illegal immigrants from the population tallies used for apportionment.
Trump called on the Commerce Department in August to begin work on “a new and highly accurate CENSUS” amid renewed scrutiny among conservatives of the 2020 count.
“People who are in our Country illegally WILL NOT BE COUNTED IN THE CENSUS,” he wrote on Truth Social.
His top aide, Stephen Miller, said in August that Democrats “rigged the 2020 census by including illegal aliens” in the count.
The Commerce Department did not respond to a request for comment about its work on the 2030 census or what steps it has taken to meet Trump’s demands for a new census, which would not affect the number of congressional seats each state has for the rest of the decade.
But the Trump administration has already begun to make its mark on the Census Bureau.
Lutnick disbanded the 2030 Census Advisory Committee in February, determining that the committee had fulfilled its mandate. He also disbanded two other advisory committees, including the National Advisory Committee on Racial, Ethnic, and Other Populations. The committee’s members included people from left-leaning groups, among other outside experts.
Last month, Lutnick also installed a new acting head of the Census Bureau, George Cook, who is concurrently serving as the undersecretary for economic affairs at the Commerce Department.
The Census Bureau under Trump also quickly stopped work on a $10 million research effort, begun during the Biden administration, to determine how best to ask LGBT-friendly questions about sexual orientation and gender identity on the census, former Census Bureau Director Robert Santos told NPR in February.
Republican lawmakers and researchers hope to avoid the mistakes of 2020 and gain greater control over preparations for the 2030 census with a few years left to course correct. Others, including House Oversight Committee Chairman James Comer (R-KY), are still pushing for answers about what went wrong in 2020.
“If you account for what I think is probably what has been weaponization, then I think that massively helps the Republican caucus in the next redistricting process,” Miller said. “I can’t imply motive, but when we spent the last 10 years looking at weaponization scheme after weaponization scheme, my head goes straight to, there was some weaponization here.”